A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 12, 18:30 UTC
[ROCm][CI] Fix config/test_config_generation.py (#29142)
Signed-off-by: charlifu <charlifu@amd.com>
Charlie Fu committed
Nov 21, 2025 at 11:12 UTC
75648b16ddce1bff02c39c6f06be62a58385ff52
1 file changed
+2
-2
docker/Dockerfile.rocm
+2
-2
@@ -7,6 +7,8 @@ FROM ${BASE_IMAGE} AS base
7
8
ARG ARG_PYTORCH_ROCM_ARCH
9
ENV PYTORCH_ROCM_ARCH=${ARG_PYTORCH_ROCM_ARCH:-${PYTORCH_ROCM_ARCH}}
10
+ENV RAY_EXPERIMENTAL_NOSET_ROCR_VISIBLE_DEVICES=1
11
+ENV RAY_EXPERIMENTAL_NOSET_HIP_VISIBLE_DEVICES=1
12
13
# Install some basic utilities
14
RUN apt-get update -q -y && apt-get install -q -y \
@@ -121,8 +123,6 @@ COPY --from=export_vllm /benchmarks ${COMMON_WORKDIR}/vllm/benchmarks
123
COPY --from=export_vllm /examples ${COMMON_WORKDIR}/vllm/examples
124
COPY --from=export_vllm /docker ${COMMON_WORKDIR}/vllm/docker
125
124
-ENV RAY_EXPERIMENTAL_NOSET_ROCR_VISIBLE_DEVICES=1
125
-ENV RAY_EXPERIMENTAL_NOSET_HIP_VISIBLE_DEVICES=1
126
ENV TOKENIZERS_PARALLELISM=false
127
128
# ENV that can improve safe tensor loading, and end-to-end time